I did some more testing and also read some more opinions here at the forum. It seems that DP67BG (and maybe other - all? - mobos with P67 chipset) only work with some specific PCI cards.
Beyond the cards cited previously, I also tested a simple Realtek-based NIC adapter, that worked, and an ancient PCI video card that gave only garbage on the screen, but no POST errors. All the cards work normally on an older computer. So far, I've tested 5 cards and only 2 worked.
Other users told their experiences with cards that don't work with DP67BG:
M-Audio Delta 1010 sound card: http://communities.intel.com/thread/22292?tstart=240
PCI RAID card based on SIL3124: http://communities.intel.com/thread/22972?tstart=90
Asus Xonar D1 sound card: http://communities.intel.com/thread/23178?tstart=0
There is an important point we may be missing. Board revision (AA number). Mine is G10491-303 (the first B3 release). I've checked on Intel site that versions -304, -305 and -306 have already been released and -400 is expected to the end of this month. So we must check if these incompatibilities affect all revisions or only some of them.
This way, I ask for the users that have PCI cards that don't work with DP67BG, to inform the AA number revisions of their motherboards.
Well, no one replied...
Updated BIOS to last version (2040). Still no change.
However, thinking about the strange behaviour with soundcards (sample rate mismatch, "cuts" in and so on) I suspect that may exist some bug related to PCI clock. Maybe the bus is not operating at the correct frequency. It appears also that the cards are running in high temperature, but I still have to check this in detail.
For the consumer, it doesn't matter if the chipset has native support to PCI or not. We bought motherboards that do have PCI slots and we just expect them to work. We assume that the motherboard manufacturer (Intel in this case) should have provided the necessary means to it if the chipset itself does not, in other words, the PCI-to-PCIe bridge chips must work.
The cited article says this situation should be a concern only for specific applications and unlikely to cause problems to the regular consumer. However, it seems it turned to be the opposite: the PCI bus only works for specific applications...
I would still like to know if this problem is isolated to Intel-branded motherboards (or maybe specific models/revisions) or if any P67/H67/H61 motherboard, from any manufacturer, will have the same problem. Because in the first case, I should RMA my board, in the second I must live with it.
I should have an answer to that question later today. I am trying an alterrnate manufacturers board, the PCIe to PCI bridge in question is the same on both boards. The designs look nearly identical. I am betting on the BIOS being the issue. I will let you know. I did not have time to wait for Intel to decide there is a problem nor the time for them to fix it, I returned the two Intel boards I bought, both had the exact same problem.
Ok, I have replaced my Intel board (DH67CLB3) with an identical competitor board. The competitor board uses the same chipset and PCIe to PCI bridge, IT8892. The GB board does not have the problem that I had with the intel board. The software now recognizes the PCI board with the PCI9050 interface chip and works normally. I don't know if it will solve your PCI problem but at least I know now that my problem was a BIOS issue.
Could you please tell which board you tested?
I'm not sure if it's a BIOS issue, maybe the Intel boards have some kind of physical defect, or project flaw that may be corrected. I remind that the "alternate power LED" only was made to work at board revision 400 that has been just released.... before that it was just a dummy conector on the board.
Anyway, the information is important since it shows that PCI support on Sandy Bridge boards is possible, the failures are not "by design".
The replacement board I used was a home run for my project, the embedded application performs better than ever and it is quite old. I am very impressed with the performance of the Sandy Bridge processor and H67 chip set. It seems that Intel has worked out all of the performance issues of the past.
Back to the PCI Issue, I am by no means an expert in this area, but while troubleshooting this issue I noticed the PCI memory mapping was quite different than with other systems. The documentation for the motherboard states the following. I don't know if its related but it seems to me that Intel may have placed the integrity of the 32GB memory map over the integrity of the PCI bus mapping.
The board utilizes 32 GB of addressable system memory. Typically the address space
that is allocated for PCI Conventional bus add-in cards, PCI Express configuration
space, BIOS (SPI Flash device), and chipset overhead resides above the top of DRAM
(total system memory). On a system that has 32 GB of system memory installed, it is
not possible to use all of the installed memory due to system address space being
allocated for other system critical functions. These functions include the following:
• BIOS/SPI Flash device (16 Mbit)
• Local APIC (19 MB)
• Direct Media Interface (40 MB)
• Front side bus interrupts (17 MB)
• PCI Express configuration space (256 MB)
• PCH base address registers PCI Express ports (up to 256 MB)
• Memory-mapped I/O that is dynamically allocated for PCI Conventional and PCI
Express add-in cards (256 MB)
The board provides the capability to reclaim the physical memory overlapped by the
memory mapped I/O logical address space. The board remaps physical memory from
the top of usable DRAM boundary to the 4 GB boundary to an equivalent sized logical
address range located just above the 4 GB boundary. Figure 8 shows a schematic of
the system memory map. All installed system memory can be used when there is no
overlap of system addresses.
Are you using a 32-bit version of Windows? I chose the 32-bit version of windows 7 for company related software compatibility. It may be this problem would not exist on 64-bit O/S.
The PCI malfunction - at least in my case - it's not a matter of OS. I had already tested an old videocard that does not give image (only garbage) right from the start, even before the OS boots. Just to be sure I've tested with 64-bit Ubuntu and the soundcard behaves exactly the same as with the 32-bit OS. I don't have a 64-bit Windows but I think the result will be the same.
I will contact Intel customer support soon and try to get some information if they are aware of the problem and expect to fix it (on a French forum, a user obtained the information about the 400 revision, that fixes the power led issue, long before the release). In the worst case I will return the board and look for a Gigabyte one.
I spent over 2 weeks and a lot of money on QUAD port PCI NIC cards in BOXDH67CLB3 board with i7-2600k CPU and 16GB RAM
The first issue:
Single card in PCI Slot 1(listed as slot 2 in BIOS), card is seen by the BIOS as inserted 32-bit card, OS does not see Ethernet controllers, tried 32 bit Linux, 32 bit windows, 64 bit linux, 64 bit windows, simply nothing.... Multiple Cards, Multiple Chipsets (Intel, DEC, Broadcom etc..)
The Second issue:
Multiple cards or single card in Slot 3 (listed as slot 4 in BIOS) the PC wont POST, get 2 Long beeps, followed by a pause then 2 Long beeps again. Manuall indecated that no Video card is present and that it`ll cont booting, which it does not.
I was searching a lot on line and couldnt find this issuse listed anywhere else, mainly tought it was the QUAD NICs, now I see its not.
I finally swapped the board with my old one, for this project, really wanted to use this CPU and this much RAM for what I`m building.
Hopefully this will help others to stay away from this board for PCI devices. I bought it specificly for 3 PCI slots, to build a 16(17 with onboard) Network Port Machine, but this did not work out. Now have to do 2 PCI QUAD and 2 PCIe Dual to make this work on my old P5Q Turbo with Q9550 and 8GB RAM.
Theres nothing wrong with the chipset, memory, or processor. I had a PCI issue with the Intel BOXDH67CLB3 with similar PCI issues, a legacy PCI board simply would not work. RMA the Intel board and got a gigabyte equivalent, the only hardware difference was the network interface, they even used the same PCIe to PCI bridge, the system has been working perfectly since then.